Orbit v0 launches a multi-agent evaluation framework on top of Inspect
j_foerst · x · 2026-07-23
Orbit v0 launches as a multi-agent evaluation framework built on @AISecurityInst’s Inspect.
- It starts with security use cases, but is intended for broader safety and capability research.
- The goal is to cut down the “weeks of infrastructure before your first result” problem by letting researchers swap environments, threat models, and topologies without rebuilding the whole stack.
Related event: Orbit v0: A New Framework for Multi-Agent Safety Eval(2 posts)→
More from coding & agent
- Atomic Agent targets local open-model use with 1,000+ skills and MCP servers — eyishazyer · 2026-07-23
- video-use edits raw footage into final.mp4 by reading transcripts, not frames — alex_verem · 2026-07-23
- Citrolabs opens ego-lite, a browser for humans and AI agents to work in parallel — citrolabs · 2026-07-23
- Alibaba open-sources open-code-review, a hybrid LLM agent for line-level code review — alibaba · 2026-07-23
- Codex feels cleaner on a blank project than inside a messy legacy codebase — Individual-Carob5593 · 2026-07-23
- LoHoSearch turns a 7.62M-entity knowledge graph into a harder benchmark for search agents — 美团技术团队 · 2026-07-23